Deformable Scene

Deformable scene reconstruction aims to accurately model and represent three-dimensional environments undergoing non-rigid transformations, a crucial challenge across diverse fields like robotics and medical imaging. Current research focuses on developing robust algorithms, often leveraging neural networks (e.g., NeRFs, Gaussian splatting) and incorporating physics-based constraints to handle complex deformations and occlusions from various input modalities (e.g., multi-view video, monocular RGB). These advancements enable improved applications such as accurate 3D human motion capture in dynamic scenes, precise robotic manipulation of deformable objects, and enhanced navigation in minimally invasive surgery.